FSI urges risk‑based AI oversight, warns against piling on new rules

The Financial Services Institute is pressing policymakers to avoid kneejerk rulemaking on artificial intelligence and instead lean on existing investor‑protection regimes, while calibrating oversight to the risk of specific use cases.

In a new white paper published Wednesday, the group argues that Regulation Best Interest and fiduciary standards already cover many conflicts and supervision obligations, with new, AI‑specific mandates warranted only when technology introduces novel harms or materially changes existing risks.

The institute also elevates investor education and literacy as a core compliance tool, paired with transparent disclosures about how AI features in advice and operations.

Low‑risk applications — such as meeting transcription, administrative automation and internal research aids — should face lighter touch requirements, the FSI white paper says, while higher‑risk tools that help make or execute investment decisions should carry tighter controls, documentation, and periodic testing with clear human accountability.

“As we continue to navigate this new AI era, it is essential that our industry has clear, practical policies and practices in place to adopt these tools responsibly and effectively,” Dale Brown, president and CEO of the Financial Services Institute, said in a statement.

Brown said AI can “streamline processes and enhance the client experience,” adding that it requires thoughtful implementation and collaboration across the industry.

FSI's latest call against duplicative regulation comes just on the heels of FINRA spotlighting generative AI and cyber‑enabled fraud in its 2026 oversight priorities. In a newly added section to its yearly report, the self-regulator for the brokerage industry explained how firms are piloting large‑language‑model tools to summarize documents and surface information from policies and client files.

Underscoring the importance of investor protection and maintaining the integrity of capital markets, FINRA urged firms to maintain governance systems that test for accuracy and bias, logs prompts and outputs, as well as ensuring supervision, communications, recordkeeping – a facet of regulation that's way past due for modernization, according to SIFMA – and fair‑dealing rules still apply when AI is in the loop.

FINRA also flagged AI agents that plan and execute tasks across systems, noting the potential for overreach, audit challenges, and sensitive‑data mishandling if firms allow models to act without adequate guardrails.

Against that backdrop, FSI’s proposed playbook focuses on practical adoption steps for independent firms and RIAs. It recommends ranking projects with a nine‑factor scoring matrix that weighs business impact, risk, time to market, data readiness, technical feasibility and ROI.

On the plumbing, a four‑stage interoperability roadmap starts with secure data exchange via APIs and event‑driven architecture, then moves to common domain and security models before enabling cross‑vendor explainability, documentation, and confidence metrics.

Case studies in the paper touted early productivity wins from AI, including a 40% cut in administrative time from automated meeting notes, a 25% increase in client coverage from automated portfolio reviews, and a 30% reduction in integration costs from standardized data taxonomies.

Bob Coppola, a group lead on the AI task force and chief technology officer at Sanctuary Wealth, said the industry needs standards that support “innovation, transparency, security and responsible use,” adding that the paper “lays the foundation for scalable, consistent AI adoption.”